By Ananya Krishnan, Operations, CineVerse Theatres.
CineVerse Theatres runs nine theatres and 12,400 seats in Hyderabad. Our dashboard problem is not producing more charts. It is getting a duty manager, film programmer, and concession lead to the same next action.
We organise the operating view around three questions:
- Scheduling: which show or screen needs attention?
- Footfall: when are arrivals building relative to showtime?
- Concessions: are transaction rate and basket size moving differently?
Start with grain, not visuals
We keep three kinds of facts separate:
| Fact table | Grain | Useful fields |
|---|---|---|
show_fact |
One row per scheduled show | Theatre, screen, film, scheduled and actual start, sellable seats, admissions, estimated screen-clear time |
entry_bucket_fact |
One row per counter and 15-minute bucket | Theatre, counter, bucket start, entries, source timestamp |
pos_line_fact |
One row per concession receipt line | Transaction ID, sale time, item, quantity, net sales, refund/void status |
Shared dimensions—such as theatre, screen, film, business date, daypart, and item—make the views comparable.
One trap is joining raw POS lines directly to shows and then summing admissions. Every receipt line can duplicate the show-level measures. We aggregate each fact separately before comparing them at a common theatre, business-date, and daypart grain.
Scheduling: seats belong to shows
The chain’s 12,400 seats describe our scale; they are not the denominator for a day’s occupancy. Each show has its own sellable capacity after holds and closures.
show occupancy = ticket admissions / sellable seats for that show
late-start delay = actual start - scheduled start
Alongside occupancy, the schedule view surfaces:
- scheduled versus actual starts;
- insufficient clear-and-clean turnaround buffers;
- the number of shows starting in the same lobby-arrival window;
- dark screens during review-worthy programming slots.
A chain-wide average is useful for a weekly review, but intervention usually happens at show or screen level. Film-scheduling research similarly treats film, screen, start time, per-show demand, and operating constraints together. An optimizer is not required for these first useful checks.
Footfall: entries are not admissions
Door counters, ticket scans, and POS transactions answer different questions.
We plot entries in 15-minute buckets alongside ticket admissions and align them with upcoming showtimes. This helps distinguish a lobby arrival surge from a show that simply has low admissions.
We deliberately do not call entries / admissions a conversion rate. Depending on the installation, counter events can include staff, re-entry, and people who do not attend a screening. Counter outages are shown as no data, never as zero footfall.
Aggregated counts are sufficient for this operating view. If a footfall source captures identifiable personal data, it needs a separate legal and privacy review; a dashboard is not a reason to collect more data.
Concessions: split spend into two drivers
Total concession revenue rises with admissions, so revenue alone is hard to interpret. We use three measures over the same theatre and time window:
transaction rate = distinct concession transactions / ticket admissions
average receipt = net concession sales / distinct concession transactions
spend per admission = net concession sales / ticket admissions
With consistent scopes and denominators:
spend per admission = transaction rate × average receipt
Illustrative example—not CineVerse data: 100 admissions, 40 concession transactions, and ₹8,800 in net sales produce:
- a 40% transaction rate;
- a ₹220 average receipt;
- ₹88 spend per admission.
If transaction rate falls while average receipt holds, we investigate arrival bunching, queues, staffing, and stock availability before drawing conclusions about the menu or pricing. If one receipt serves a group, transaction rate is a proxy—not a measure of unique buyers.
“Net sales” must have one documented meaning across theatres, including the treatment of taxes, refunds, and voids.
Make the dashboard actionable
We reuse one dashboard with theatre, business-date, film, and daypart filters rather than maintaining separate copies for each location. Every panel displays the last successful source timestamp.
Operational alerts focus on things someone can act on:
- stale or failed feeds;
- actual starts beyond the agreed late threshold;
- counter outages;
- arrival surges against the staffing plan;
- queue or stockout signals, where those are captured.
Low occupancy alone goes to programming review rather than automatically paging a duty manager.
Refresh frequency follows the source. Refreshing a panel every 30 seconds does not improve a 15-minute extract. Business dates use the Asia/Kolkata timezone and an explicit rule for post-midnight shows.
Our basic rule is that every operational KPI needs an owner, grain, denominator, freshness expectation, and next action. If a panel cannot tell someone what to check next, it probably belongs in an exploratory report rather than the operations dashboard.
Sources and further reading
- Demand-driven scheduling of movies in a multiplex — Eliashberg et al.
- Optimisation approaches and challenges of personnel deployment for movie exhibitors — Brühl and Eigler
- The economics of the operation — Independent Cinema Office
- Understand star schema — Microsoft
- Grafana dashboard best practices
- India’s Digital Personal Data Protection Act, 2023 and the PIB summary of the DPDP Rules, 2025
Top comments (0)